Is diffraction an increasing worry? Or, what's wrong with more megapixels?

If the diameter of a circle of light equals the size of 2x2 photosites, we call it (Bayer-arrayed) 16 Mpixels. If the diameter of a circle of light equals the size of 4x4 photosites, I call it (Bayer-arrayed) 4 Mpixels. If the diameter of a circle of light equals the size of 16x16 photosites, I call it (Bayer-arrayed) 1 Mpixels. Is that "bone-headed" thinking ? What am I doing wrong there ?

Are you talking about light that exists outside of the diameter of such circles (yet inside the square-shaped areas of the NxN photo-site arrays that I am describing) ? Assuming an (unlikely) 100% Fill Factor, that represents (at most) 21.46% of the square area. Is that the "missing link" ?

I have seen that identity mentioned in various writings. I have also seen it stated many times that such an identity represents only a rough and often not representative approximation of what occurs. Am curious as to the nature of the derivation of that identity. Do you have any related information or references that derive and show the nature and validity of that identity ? Am sincerely interested.